Data-driven models for predicting fire and explosion-related properties have been improved greatly in recent years using machine-learning algorithms. However, choosing the best machine learning approach is still a challenging task. Therefore, in this study, the predictability comparisons have been made with the different machine learning methods used to model the MIT for iron dust. The MIT of iron dust was determined using the Godbert-Greenwald furnace for seventy unique combinations of dispersion pressures and dust concentrations. The data has been divided into 'Training Set' and 'Testing Set'. The implementation and efficacy of machine learning and statistical approaches have been demonstrated through real-time experimental results. The support vector machines (SVM) regression models were trained with various kernel functions to enhance the performance of the resultant model. The cubic kernel function was found suitable for training SVMs. Besides, a feed-forward artificial neural network with the backpropagation algorithm and a polynomial surface fit model have also been developed to predict the MIT. For statistical phenomena, such as MIT, predictive modelling based on real-time experimental data is critical. If an accurate estimate of the combustible dust's minimum ignition temperature is confirmed, it is possible to ensure that the temperatures of the surrounding hot surfaces do not rise to that level, preventing an explosion. An overall comparison of predictive models has been given with unseen test data set. All the trained models yielded comparable results with unseen test data set. However, the SVM model with Bayesian optimizer approach can effectively assess the risk of ignition based on dust MIT under the influence of dispersion pressure and dust cloud concentration among all the approaches adopted in this study. © 2021 Institution of Chemical Engineers